Re: R on riscv64 (Re: building fewer packages or reduced functionality on i386)
Richard Lewis <richard.lewis.debian-gM/[email protected]>
| Newsgroups | gmane.linux.debian.devel.general,gmane.linux.debian.riscv |
|---|---|
| Message-ID | <[email protected]> |
Rene Engelhard <[email protected]> writes: > Hi, > > Am 19.08.26 um 23:57 schrieb Charles Plessy: >> Le Sat, Aug 15, 2026 at 07:30:37PM +0100, Richard Lewis a écrit : >>> >>> I think you should ask yoursekf: what is the ask of upstream? >> […] >>> bluntly, from the start, i gather R doesnt support riscv64 at all: so >>> why should they care? >> >> Thanks Richard for your comments, >> >> What we uncovered by building on riscv64 may be either a niche problem >> that upstream does not have to feel responsible for, or an existential >> problem that usptream has to engage with years ahead with hard work: >> what if the next mainstream architecture on Win, Apple or Linux also >> does not have NaN payload propagation like riscv64 today? >> > LibreOffice is in the same boat. > > > FWIW, that sounds like the same as/related to > > https://salsa.debian.org/libreoffice-team/libreoffice/libreoffice/-/blob/master/libreoffice-calc.templates?ref_type=heads > > shown only on risvc64, > cf. https://salsa.debian.org/libreoffice-team/libreoffice/libreoffice/-/blob/master/libreoffice-calc.postinst.in?ref_type=heads Have you also tried to discuss this issue with whichever upstream project (glibc? kernel?) defines the NaN propogation behaviour?